Iron stable isotopes track pelagic iron cycling duringa subtropical phytoplankton bloomMichael J. Ellwooda,1, David A. Hutchinsb, Maeve C. Lohanc, Angela Milnec, Philipp Nasemannd, Scott D. Noddere,Sylvia G. Sanderd, Robert Strzepeka, Steven W. Wilhelmf, and Philip W. Boydg,2

aResearch School of Earth Sciences, Australian National University, Canberra, ACT 2601, Australia; bMarine and Environmental Biology, Department ofBiological Sciences, University of Southern California, Los Angeles, CA 90089; cSchool of Geography, Earth and Environmental Sciences, University ofPlymouth, Plymouth PL4 8AA, United Kingdom; dMarine and Freshwater Chemistry, Department of Chemistry, University of Otago, Dunedin 9054, NewZealand; eNational Institute of Water and Atmospheric Research, Wellington 6021, New Zealand; fDepartment of Microbiology, The University of Tennessee,Knoxville, TN 37996; and gNational Institute of Water and Atmospheric Research Centre for Chemical and Physical Oceanography, Department of Chemistry,University of Otago, Dunedin 9012, New Zealand

Edited by Edward A. Boyle, Massachusetts Institute of Technology, Cambridge, MA, and approved December 1, 2014 (received for review November 11, 2014)

The supply and bioavailability of dissolved iron sets the magnitudeof surface productivity for ∼40% of the global ocean. The redoxstate, organic complexation, and phase (dissolved versus particu-late) of iron are key determinants of iron bioavailability in themarine realm, although the mechanisms facilitating exchange be-tween iron species (inorganic and organic) and phases are poorlyconstrained. Here we use the isotope fingerprint of dissolved andparticulate iron to reveal distinct isotopic signatures for biologicaluptake of iron during a GEOTRACES process study focused ona temperate spring phytoplankton bloom in subtropical waters.At the onset of the bloom, dissolved iron within the mixed layerwas isotopically light relative to particulate iron. The isotopicallylight dissolved iron pool likely results from the reduction of par-ticulate iron via photochemical and (to a lesser extent) biologicallymediated reduction processes. As the bloom develops, dissolvediron within the surface mixed layer becomes isotopically heavy,reflecting the dominance of biological processing of iron as it isremoved from solution, while scavenging appears to play a minorrole. As stable isotopes have shown for major elements like nitro-gen, iron isotopes offer a new window into our understanding ofthe biogeochemical cycling of iron, thereby allowing us to disen-tangle a suite of concurrent biotic and abiotic transformations ofthis key biolimiting element.

iron isotopes | marine biogeochemical cycles | trace metals |phytoplankton blooms | GEOTRACES

Springtime phytoplankton blooms are major contributors tothe drawdown of carbon dioxide (CO2) from the atmosphere

and its sequestration into the ocean’s interior (1, 2). In thecontext of the ocean’s iron (Fe) biogeochemical cycle, springblooms represent a transition from early season production,fueled largely by new Fe from underlying waters or lateral supply(3), to postbloom conditions where primary production is mainly(i.e., ∼90%) supported by an efficient Fe recycling loop betweenbiogenic particulates and the dissolved Fe pool (3). Photochemicalreduction and biological processing of inorganic, complexed, andparticulate Fe significantly enhances Fe bioavailability (4–7);however, our understanding of the mechanisms, timing, and ratesof Fe exchange between pools (dissolved, lithogenic, and bio-genic) and redox species (FeII and FeIII) during the onset anddevelopment of a phytoplankton bloom is limited. Indeed, thetransient nature of many of these processes makes it difficult toquantify their influence on the biogeochemical cycling of Fe.Iron isotope ratios (56Fe/54Fe) are a promising tool because

isotope fractionation can occur upon transformation of Feredox species (8), particulate dissolution (9), scavenging (10),precipitation (11), and biological uptake by phytoplankton (12).To date, a limited number of open ocean Fe isotope studies havebeen published (10, 12–14), with few combining both dissolvedand particulate data to trace exchanges between various Fepools. Here, we present Fe isotope data from two GEOTRACES

(www.geotraces.org) process voyages (2008 and 2012) designedto study temporal changes in the biogeochemical cycling of Fe atthe same locality in subtropical waters (38°S−39°S, 178°W−180°W) within the mesoscale eddy field east of New Zealand(3). We first present in situ results for dissolved and particulateFe (DFe and PFe) cycling during the annual spring bloom, fol-lowed by the findings from a shipboard 700-L mesocosm in-cubation experiment, and then a conceptual model outlining thekey chemical and biological processes involved in Fe isotopefractionation.

Results and DiscussionAcross the two voyages, we identified three distinct stages as-sociated with the progression of the annual spring bloom. Stage Iis characterized by low Net Primary Productivity (NPP) (1.54μmol C·L−1·d−1), low chlorophyll a (Chl) concentrations, lowbiomass (Fig. 1 and Fig. S1), and relatively homogenous ni-trate and DFe profiles between 0 m and 250 m (Fig. 2A). This isindicative of a system that has been reset by turbulent mixing andconvective overturning during winter (15) and primed, environ-mentally, for phytoplankton to bloom. Stage II is characterizedby the initial development of a diatom-dominated bloom, in-creasing rates of NPP (6.15 μmol C·L−1·d−1) and phytoplanktonand grazer biomass (Fig. 1 and Fig. S1), resulting in partial

Significance

The supply and bioavailability of dissolved iron sets the mag-nitude of surface productivity for approximately 40% of theglobal ocean; however, our knowledge of how it is transferredbetween chemical states and pools is poorly constrained. Herewe utilize the isotopic composition of dissolved and particulateiron to fingerprint its transformation in the surface ocean byabiotic and biotic processes. Photochemical and biological re-duction and dissolution of particulate iron in the surface oceanappear to be key processes in regulating its supply and bio-availability to marine biota. Iron isotopes offer a new windowinto our understanding of the internal cycling of Fe, therebyallowing us to follow its biogeochemical transformations in thesurface ocean.

Author contributions: M.J.E., D.A.H., S.D.N., R.S., S.W.W., and P.W.B. designed research; M.J.E.,M.C.L., A.M., P.N., S.D.N., S.G.S., R.S., S.W.W., and P.W.B. performed research; M.J.E.contributed new reagents/analytic tools; M.J.E., D.A.H., A.M., P.N., S.G.S., R.S., and S.W.W.analyzed data; and M.J.E., D.A.H., M.C.L., S.G.S., S.W.W., and P.W.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. Email: Michael.Ellwood@anu.edu.au.2Present address: Institute for Marine and Antarctic Studies, University of Tasmania, Hobart,TAS 7005, Australia.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1421576112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1421576112 PNAS | Published online December 22, 2014 | E15–E20

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

drawdown of nitrate and DFe within the mixed layer (Figs. 1and 2B). Stage III is characterized by elevated NPP (10.7 μmolC·L−1·d−1), higher Chl concentrations/biomass (Fig. S1) and acorresponding biological depletion of nitrate and DFe withinand immediately below the mixed layer (Figs. 1 and 2C andFig. S1) (3, 16).A time-dependent change in the δ56Fe composition of DFe

and PFe is observed across stages I to III (Fig. 2 and Fig. S2).During stage I, the δ56Fe composition of DFe and PFe within theeuphotic zone was different, with lighter δ56Fe values for DFe(Δ56FePFe-DFe = 0.28‰) relative to PFe (Fig. 2D). The dissolvedδ56Fe composition varied vertically whereby δ56Fe values in-creased with depth (100–300 m), even though nitrate and theDFe profiles were homogenous (5.10 ± 0.19 μmol·L−1 and 0.38 ±0.02 nmol·kg−1, respectively). At 300 m, the δ56Fe compositionof DFe and PFe was isotopically the same (0.04 ± 0.09‰ and0.08 ± 0.01‰, respectively) and consistent with an inferred

lithogenic provenance of coastally derived particulates (Fig. 2Dand Fig. S2) (3, 12, 16, 17).During stage I, there are two candidate processes that could

lead to an isotopically light DFe pool within the euphotic zone:photochemical and biological reduction of PFe, the latter viaacidic phagocytosis upon ingestion by protozoan grazers (18–20).The key process required for δ56Fe fractionation is the reductionof FeIII to FeII and its subsequent release into solution; δ56Fefractionation associated with proton-promoted dissolution oflithogenic Fe (e.g., goethite and hematite), as might occur in thedigestive gut of grazers, is likely to be less (9, 21) compared withδ56Fe fractionation associated with photochemical reduction oflithogenic Fe. It should also be noted that acidic and enzymaticdigestion of PFe by grazers may also promote Fe reduction andsolubilization (20), but it is usually followed by exposure to al-kaline conditions, which leads to reoxidization before egestion(20). If a portion of this reduced, isotopically light Fe is taken upby the grazer, then this would lead to an isotopically heavier Fe

Fig. 1. Satellite-derived chlorophyll concentrationsfor stages I, II, and III of the subtropical springphytoplankton bloom. Diamond represents thesampling site. Gray areas represent cloud coverduring the satellite pass-over. MODIS Aqua satellitedata obtained from ERDDAP and plotted usingGeneric Mapping Tools (46).

0 5 10 15 20 25

NO3- mol L-1

0 0.5 1 1.5

DFe (nmol kg-1)

750

625

500

375

250

125

0

Depth (m

)

0 5 10 15 20 25

0 0.5 1 1.5

750

625

500

375

250

125

0

Dep

th (m

)

0 5 10 15 20 25

0 0.5 1 1.5

FeNO3

-

Stage IIIStage IIStage I

-0.4 0 0.4 0.856Fe (‰)

-0.4 0 0.4 0.8

56Fe (‰)

750

625

500

375

250

125

0

Depth (m

)

-0.4 0 0.4 0.8

-0.4 0 0.4 0.8

750

625

500

375

250

125

0

Dep

th (m

)

-0.4 0 0.4 0.8

-0.4 0 0.4 0.8

PFe DFeDFe PFePFeDFe

Stage IIIStage IIStage I

A B C

D E F

Fig. 2. DFe and δ56Fe depth profiles. (A−C) Depth profiles of DFe and dissolved nitrate concentration across the three stages of the annual spring phyto-plankton bloom. (D−F) The δ56Fe results for dissolved, suspended (diamonds), and sinking particulates (upside-down triangles) across the three stages of thephytoplankton bloom. Error bars represent either 2 SDs for multiple sample extraction and isotope separations or 2 SEs of instrument precision for a singlesample extraction and isotope separation.

E16 | www.pnas.org/cgi/doi/10.1073/pnas.1421576112 Ellwood et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

composition of the remaining Fe pool upon reoxidation and lossvia egestion. At this stage, we cannot fully disentangle thecontributions of these two processes (photochemical versusgrazer-mediated biological processing of lithogenic Fe) to theisotopically light dissolved Fe pool, but note from the in-formation available that the photochemical reduction rate islikely to be two to three times higher than that of grazer-medi-ated Fe processing during stage I when grazer biomass andbacterial abundance were low (Table S1 and Fig. S1). Clearly,though, more work will be needed to distinguish between pho-tochemical and biological effects on particulate iron dissolutionand isotopic fractionation. These multiple lines of evidence(relationship between DFe and PFe, and isotopic signatureswith depth) and, in particular, the low dissolved δ56Fe valueswithin the euphotic zone are consistent with the release of iso-topically light Fe from lithogenic particulate material (22–25).During the bloom onset (stage II), the δ56Fe compositions of

DFe and PFe within the mixed layer are the same within error(Δδ56FePFe-DFe = 0.05‰), indicating a biological influence onδ56Fe fractionation (Fig. 2E). This is evident in size-fractionatedplankton samples (0.2–20 μm and >20 μm) within the mixedlayer, with the 0.2- to 20-μm size fraction being 0.15‰ lighterthan the >20-μm size fraction (Fig. 3A). Pools of Fe associatedwith small phytoplankton are known to turn over on a timescaleof hours (26); thus δ56Fe fractionation in the 0.2- to 20-μm sizefraction likely reflects fractionation associated with the rapidrecycling of Fe between the DFe pool and biogenic componentswithin the PFe pool. In addition, small (e.g., Synechococcus) andlarge phytoplankton (e.g., the diatom Asterionellopsis glacialis)

may also fractionate δ56Fe to differing degrees, although we didnot see δ56Fe fractionation between differing size classes in ourmesocosm experiment (see below). Below the mixed layer, theδ56Fe composition of DFe is lighter than PFe (Δ56FePFe-DFe =∼0.2‰ at 100 m) (Fig. 2E), which is still characteristic of a sys-tem reset by winter mixing (15), even though DFe levels are∼0.1 nmol·kg−1 lower than during stage I.At the peak of the bloom (stage III), the δ56Fe composition of

DFe within the mixed layer is heavier than the δ56Fe composi-tion of particulate material (Δ56FePFe-DFe = −0.26‰), consis-tent with isotope fractionation during biological uptake (Fig. 2F).Below the mixed layer, the δ56Fe composition of DFe is alsoheavier than PFe and is linked to the depletion of DFe (Fig. 2C);the concentration of DFe at 100 m is ∼0.22 nmol·kg−1 lower thanduring bloom stage I.The overall change in Δ56FePFe-DFe across bloom stages I to

III is −0.54‰ and is indicative of δ56Fe fractionation mainlyassociated with DFe uptake by small phytoplankton (12). Thechanges observed in the δ56Fe composition of DFe and PFeduring the evolution of the bloom are supported by changes inthe particulate Fe to aluminum (Fe:Al) ratio of particulatematter and the percentage of biogenic Fe to the total PFe pool;both parameters increase across stages I to III (Fig. 3B).To further interpret our field results, we conducted a 700-L

phytoplankton mesocosm experiment, using water collectedduring bloom stage I (Fig. 4). During this time-course incubationstudy, fluorescence (F0), as an indicator of Chl biomass, in-creased while nutrients (NO3 and Si) and DFe were drawn downas a phytoplankton bloom developed over an 8-d period (Fig. 4).The bloom-forming diatom Asterionellopsis glacialis dominatedbiomass after day 3, which is consistent with our field resultswhere this diatom species was also dominant (3, 16). In contrastto our field results, in the mesocosm experiment, no significantvariations in the δ56Fe composition of DFe or size-fractionated(0.2 μm to 2 μm, 2 μm to 20 μm, and >20 μm) PFe were observed(Fig. 4F). The differences between our field and mesocosm δ56Feresults can be reconciled in the following ways: First, during thein situ field experiment, the Fe uptake was dominated by the 0.2-to 2-μm and 2- to 20-μm size classes (3), whereas DFe uptake inthe mesocosm experiment was dominated by the >20-μm sizeclass (Fig. 4D); Second, we note that the fe ratio (the ratio of newFe uptake versus total uptake of new and recycled iron) declinedfrom ∼0.6 during stage II to ∼0.1 during stage III of the in situphytoplankton bloom. Because small phytoplankton dominateDFe drawdown and recycling in the in situ bloom (3) and largediatoms dominate DFe and nutrient drawdown in the mesocosmexperiment (Fig. 4), the likely driver of the observed changesin δ56Fe composition of DFe and PFe for the in situ phyto-plankton is the uptake and regeneration of Fe by small phyto-plankton (e.g., cyanobacteria) along with the export of biogeniciron to depth (16). Of course, export does not occur in themesocosm experiment as it is a closed system. In other words,biological δ56Fe fractionation associated with the in situ fieldexperiment is likely to be coupled to the frequency with whichFe has cycled through the “ferrous wheel” by the microbialcommunity and the amount of biogenic iron that is exportedfrom the mixed layer (27, 28).Scavenging and the precipitation of DFe also result in δ56Fe

fractionation (9, 11, 29). The contribution of this particle-mediated δ56Fe fractionation was explored on a third voyage in2011 by following changes in δ56Fe for DFe as it is lost from so-lution from a constant hydrothermal supply source of DFe andPFe into subtropical waters (Fig. 5A). We note that there arecaveats associated with this approach, such as the potential for theformation of multiple particulate Fe phases with differing isotopefractionation factors (11); in particular, phases formed under ki-netic control have a different δ56Fe composition comparedwith phases formed under equilibrium control (30). However,

Fig. 3. Size-fractionated PFe and δ56Fe depth profiles. (A) Depth profiles ofsize-fractionated (0.2–20 μm and >20 μm) PFe concentration and δ56Fe. (B)PFe isotope versus Fe:Al ratio for suspended particulate matter across stagesI to III along with the percentage of biogenic Fe for PFe. The percentage ofbiogenic Fe is iron is based on excess PFe relative to the lithogenic Fe:Al ratioof 0.18 (16). Error bars are the same as in Fig. 2.

Ellwood et al. PNAS | Published online December 22, 2014 | E17

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

our approach is justified as it represents Fe isotopic fractionationunder the relevant marine conditions (i.e., well-oxygenatedwaters at seawater pH) for DFe loss from solution by scavengingand/or mineral precipitation under abiotic conditions withinthe deeper water column (9). As DFe was lost from solution, itsδ56Fe composition increased from ∼0.07‰ to 1.73‰. Usingthese data we obtained a fractionation factor of −0.67‰ (Fig. 5B),which is similar to the change in Δ56FePFe-DFe (−0.54‰) acrossbloom stages I to III for our field study and within the range forFeIII loss from solution (Table S1). However, in our mesocosmexperiment, the percentage of Fe bound to the surface of theparticulate material decreased from 60−80% at the start of theexperiment to 20–40% as the mesocosm phytoplankton bloompeaked (Fig. 4). Thus, phytoplankton were actively taking up andretaining Fe. Likewise, the biological uptake of DFe duringstages II and III matches that of the observed water columndecrease in the mixed layer DFe inventory (3); thus the overallchange in Δδ56FePFe-DFe across stages I to III appears to be as-sociated with biological-induced isotope fractionation and notDFe scavenging. Below the euphotic zone, Fe release and scav-enging associated with the remineralization of sinking organicmatter (31) will influence the δ56Fe composition of DFe and PFe.

The candidate process(s) put forward to explain the spatialand temporal trends in our δ56Fe results are highlighted ina conceptual diagram (Fig. 6). At a depth of 300 m, during stageI, the two processes leading to δ56Fe fractionation are de-sorption/dissolution and sorption/scavenging of PFe and DFe,respectively (Fig. 6B and Table S2). In the euphotic zone, thedominant processes leading to δ56Fe fractionation are likely tobe reductive dissolution of detrital/lithogenic Fe (photochemi-cally or biologically induced) along with desorption/dissolutionand sorption/scavenging processes for PFe and DFe, re-spectively. During stages II and III, biological uptake of DFe islikely to dominate δ56Fe fractionation within the euphotic zoneas DFe is taken up by phytoplankton.Our results show that Fe cycling during the annual spring

phytoplankton bloom in subtropical waters, east of New Zea-land, is dynamic with photochemical reduction and biologicalprocessing of PFe appearing to play important roles in cycling Febetween the particulate and dissolved pools before bloom onset,after which the biological processing of DFe dominates (32–34).In low-Fe environments (e.g., the Southern Ocean, the southwestPacific, and Equatorial Pacific), diel variations in the δ56Fecomposition of the DFe pool might be expected as a result of

Fig. 4. DFe and PFe results for the large incubation bag mesocosm experiment. (A) Fluorescence (F0) versus time for the stable and radioactive Fe bags. Theincrease in F0 is consistent with an increase in plankton biomass as time progresses. (B) Drawdown of silicate and nitrate versus time for the stable Fe bag. (C)Drawdown of DFe concentration versus time for the stable and radioactive Fe bags determined by flow injection analysis and solvent extraction (see SIMethods). (D) Size-fractionated PFe concentrations for the stable bag. (E) Ratio of surface-absorbed Fe versus total PFe for size-fractionated particulatesamples labeled with radioactive 55Fe. The symbols are the same as in D. (F) Size-fractionated δ56Fe data for PFe and δ57/56Fe for DFe (purple triangles) for thestable Fe bag. The symbols are the same as in D. Error bars are the same as in Fig. 2.

E18 | www.pnas.org/cgi/doi/10.1073/pnas.1421576112 Ellwood et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

both photochemical interactions with particulate material andduring biological processing (i.e., Fe recycling by grazers, viruses,and heterotrophic bacteria); however, the challenge is to extractthis information, because determining the δ56Fe compositionDFe species at low concentrations (<0.1 nmol·L−1) is nontrivial.The present study shows that iron isotopes are a valuable di-agnostic tool to trace the photochemical, abiotic, and biologicaltransformation of DFe and PFe and will form an important newcomponent of future studies of the biogeochemical cycling of thiskey limiting nutrient in the ocean.

MethodsSample Collection. Surface seawater was either collected from a depth of∼5 m using a trace-metal-free pump system (Almatec SL20) (35) or usingacid-cleaned, 5-L Teflon-coated externally sprung Niskin bottles, attached toan autonomous rosette (Model 1018; General Oceanics). Seawater samplesfor DFe concentration and isotope measurements were filtered throughacid-cleaned 0.2-μm capsule filters (Supor AcroPak 200; Pall) and acidifiedto pH 1.8 with Teflon-distilled nitric acid.

Particulate trace metal samples were collected in situ onto acid-leached0.2-μm polycarbonate (142-mm diameter) filters (Nucleopore Whatman)using two large volume pumps (McLane Research Laboratories), deployed atvarious water depths. At a few stations, acid-leached 20-μm polycarbonatefilters (Sterlitech) were also fitted to the filter stack so that two size classeswere obtained: 0.2–20 μm and >20 μm. Sinking cells and particles wereintercepted using surface-tethered, free-drifting MULTI-trap sediment trapsdeployed at 100-, 150-, and 200-m depths, which were trace metal-cleanedand preserved using a chloroform salt brine (35–38).

Hydrothermally influenced seawater samples were collected in 2011 ad-jacent to the Brothers underwater volcano (34°52′18.6 S, 179°03′19.8 E;northwest vent depth ∼1,455 m) located along the Tonga−Kermadec arcsystem (39, 40) (SI Text and Fig. S3).

The large mesocosm experiment involved filling two acid-cleaned, 1,000-Lnylon reinforced polyethylene bags (Scholle) with filtered and unfilteredsurface seawater. Initially, the bags were filled with ∼350 L of 0.2-μm(Acropak; Pall) filtered seawater. The bags were then spiked with eitherradioactive 55Fe or stable Fe such that the final dissolved Fe concentrationwas raised by 0.2 nmol·L−1 to 0.45–0.5 nmol·L−1. The added Fe was thenallowed to equilibrate with the natural organic ligands for an 8-h period.Before dawn, each bag was then filled with unfiltered seawater containingthe natural phytoplankton community to a volume of ∼700 L. Bag tem-peratures were maintained at in situ temperature by flowing surface sea-water around each incubation bag. Each bag was shaded to 50% of incidentradiation. Time-course samples for each bag were collected periodically (6-to 24-h periods) for DFe, size-fractionated (0.2–2 μm, 2–20 μm, and >20 μm)particulate trace metals, DFe and PFe isotopes, and nutrients. Dissolved Fefor the radioactive 55Fe or stable Fe bags were determined by flow injectionanalysis with chemiluminescence detection of Fe using luminol followingtrace element preconcentration on to the Toyopearl AF-Chelate-650 M resin(Tosoh Bioscience) (41, 42).

Fig. 5. Influence of particulate scavenging and mineral precipitation on DFeand δ56Fe fractionation. (A) Depth profiles of DFe concentration and δ56Fe forsamples collected adjacent to the Brothers underwater volcano. Error bars arethe same as in Fig. 2. (B) Open and closed system Rayleigh fractionation mod-eling (47) of δ56Fe values using A for samples collected adjacent to the Brothersunderwater volcano, where F is the fraction of DFe remaining relative to a DFeconcentration of 8.31 nmol·kg−1. The closed system model produces an αscav =−0.67‰ while the open system model produces an αscav = −1.62.

Fig. 6. Cartoon highlighting the various pathways that can lead to δ56Fefractionation. (A) Stage I, depth of 300 m. (B) Stage I, euphotic zone. (C) Stages IIand III, euphotic zone. For simplicity, no differentiation between inorganic Feand Fe complexation to natural organic ligands was made; rather, we treatedinorganic Fe and organically complexed Fe as one group. Water column mea-surements from both the 2008 and the 2012 voyages indicate that the majority(>90%) of DFe was complexed to high-affinity Fe-binding ligands (FeL) (3).

Ellwood et al. PNAS | Published online December 22, 2014 | E19

EART

H,A

TMOSP

HER

IC,

ANDPL

ANET

ARY

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

The background PFe concentration at the start of the mesocosm experi-ment was estimated to be between 0.6 nmol·L−1 and 0.8 nmol·L−1, which isconsistent with the measured concentration range PFe within the mixedlayer (0.77 nmol·L−1 to 1.87 nmol·L−1).

Sample Analysis. Sediment trap and particulate samples for trace elementand δ56Fe determination were thawed and processed using the acid digestionprotocol of Eggimann and Betzer (43) as described by Ellwood et al. (16).

The δ56Fe composition of Fe was made on samples purified using theanion exchange procedure described by Poitrasson and Freydier (44). Beforepurification, DFe samples were preconcentrated by dithocarbamate extrac-tion (16). Iron isotopes were determined using a Neptune Plus multicollectorInductively Coupled Plasma Mass Spectrometer (ICPMS) (Thermo Scientific)with an APEX-IR introduction system (Elemental Scientific) and with X-typeskimmer cones. Samples were measured in high-resolution mode with 54Crinterference correction on 54Fe and with instrumental mass bias correctionusing nickel (44). Sample 56Fe/54Fe ratios are reported in delta notation (‰)relative to the IRMM-014 Fe isotope reference material [Institute for ReferenceMaterials and Measurements (IRMM)] using the standard-sample-standardbracketing technique where δ56Fe = [(56Fe/54Fe)sample/(

56Fe/54Fe)IRMM-014 – 1]·1,000.The overall sample processing and instrumental error for dissolved and

particulate Fe samples ranged between ±0.05‰ and ±0.22‰ (2σ). The δ56Fe

values obtained for the GEOTRACES standards GSI and GDI and standardreference materials, BCR-2 and NOD-A-1, were within the range of publishedvalues (Table S3) for the GEOTRACES intercalibration study (45). Our particu-late and dissolved δ56Fe measurements were correlated to δ57Fe with a δ57Fe/δ56Fe slope of 1.50 ± 0.03 (± std. error, n = 147, P < 0.001), which is within errorof the theoretical mass-dependent fractionation slope of 1.47, except for theDFe samples for the large mesocosm experiment; hence we express thesevalues as δ56/57Fe because of an interference on mass 54Fe. Unfortunately, theremaining sample volume was not enough to repeat the extraction process.

ACKNOWLEDGMENTS. We thank the Captain and crew of the RV Tangaroa forassistance at sea, and Sarah Searson and Lisa Northcote [both from NationalInstitute of Water and Atmospheric Research (NIWA)] for sediment trap andMcLane pump deployments and sample processing. We are grateful to Ed Boylefor sharing GEOTRACES standards GSI and GDI with us. We thank two reviewersand the editor for their constructive reviews. This research was supported by theNew Zealand Foundation for Research, Science and Technology Coasts andOceans Outcome-Based Investment (CO1X0501, now NIWA Core funding), theAustralian Research Council (DP110100108, DP0770820, and DP130100679), USNational Science Foundation (OCE 0825405/0825319, awarded to D.A.H. andS.W.W.), Royal Society of New Zealand Marsden Fund (U001117, awardedto S.G.S.), and the National Environmental Research Council (NERC NE/H004475/1, awarded to M.C.L.).

1. Falkowski PG, Raven JA (2007) Aquatic Photosynthesis (Princeton Univ Press, Prince-ton, NJ), 2nd Ed.

2. Mahadevan A, D’Asaro E, Lee C, Perry MJ (2012) Eddy-driven stratification initiatesNorth Atlantic spring phytoplankton blooms. Science 337(6090):54–58.

3. Boyd PW, et al. (2012) Microbial control of diatom bloom dynamics in the open ocean.Geophys Res Lett 39(18):L18601.

4. Barbeau K, Rue EL, Bruland KW, Butler A (2001) Photochemical cycling of iron in thesurface ocean mediated by microbial iron(III)-binding ligands. Nature 413(6854):409–413.

5. Finden DAS, Tipping E, Jaworski GHM, Reynolds CS (1984) Light-induced reduction ofnatural iron(III) oxide and its relevance to phytoplankton. Nature 309(5971):783–784.

6. Sunda WG, Huntsman SA (1997) Interrelated influence of iron, light and cell size onmarine phytoplankton growth. Nature 390(6658):389–392.

7. Barbeau K, Moffett JW, Caron DA, Croot PL, Erdner DL (1996) Role of protozoangrazing in relieving iron limitation of phytoplankton. Nature 380(6569):61–64.

8. Welch SA, Beard BL, Johnson CM, Braterman PS (2003) Kinetic and equilibrium Feisotope fractionation between aqueous Fe(II) and Fe(III). Geochim Cosmochim Acta67(22):4231–4250.

9. Skulan JL, Beard BL, Johnson CM (2002) Kinetic and equilibrium Fe isotope fraction-ation between aqueous Fe(III) and hematite. Geochim Cosmochim Acta 66(17):2995–3015.

10. John SG, Adkins J (2012) The vertical distribution of iron stable isotopes in the NorthAtlantic near Bermuda. Global Biogeochem Cycles 26(2):GB2034.

11. Beard BL, Johnson CM (2004) Fe isotope variations in the modern and ancient Earthand other planetary bodies. Rev Mineral Geochem 55(1):319–357.

12. Radic A, Lacan F, Murray JW (2011) Iron isotopes in the seawater of the equatorialPacific Ocean: New constraints for the oceanic iron cycle. Earth Planet Sci Lett 306(1-2):1–10.

13. Lacan F, et al. (2008) Measurement of the isotopic composition of dissolved iron in theopen ocean. Geophys Res Lett 35(L2):L24610.

14. Conway TM, John SG (2014) Quantification of dissolved iron sources to the NorthAtlantic Ocean. Nature 511(7508):212–215.

15. Tagliabue A, et al. (2014) Surface-water iron supplies in the Southern Ocean sustainedby deep winter mixing. Nat Geosci 7(4):314–320.

16. EllwoodMJ, et al. (2014) Pelagic iron cycling during the subtropical spring bloom, eastof New Zealand. Mar Chem 160:18–33.

17. Beard BL, Johnson CM, Von Damm KL, Poulson RL (2003) Iron isotope constraints onFe cycling and mass balance in oxygenated Earth oceans. Geology 31(7):629–632.

18. Wells ML, Mayer LM (1991) The photoconversion of colloidal iron oxyhydroxides inseawater. Deep Sea Res Part A 38(11):1379–1395.

19. Barbeau KA, Moffett JW (1998) Dissolution of iron oxides by phagotrophic protists:Using a novel method to quantify reaction rates. Environ Sci Technol 32(19):2969–2975.

20. Hutchins DA, Bruland KW (1994) Grazer-mediated regeneration and assimilation ofFe, Zn and Mn from planktonic prey. Mar Ecol Prog Ser 110(2-3):259–269.

21. Wiederhold JG, et al. (2006) Iron isotope fractionation during proton-promoted, li-gand-controlled, and reductive dissolution of Goethite. Environ Sci Technol 40(12):3787–3793.

22. Sulzberger B, Laubscher H (1995) Reactivity of various types of iron(III) (hydr)oxidestowards light-induced dissolution. Mar Chem 50(1-4):103–115.

23. Borer P, Sulzberger B, Hug SJ, Kraemer SM, Kretzschmar R (2009) Photoreductivedissolution of iron(III) (hydr)oxides in the absence and presence of organic ligands:Experimental studies and kinetic modeling. Environ Sci Technol 43(6):1864–1870.

24. Waite TD, Morel FMM (1984) Photoreductive dissolution of colloidal iron oxide: Effectof citrate. J Colloid Interface Sci 102(1):121–137.

25. Wells ML, Mayer LM, Donard OFX, de Souza Sierra MM, Ackelson SG (1991) Thephotolysis of colloidal iron in the oceans. Nature 353(6341):248–250.

26. Poorvin L, Rinta-Kanto JM, Hutchins DA, Wilhelm SW (2004) Viral release of iron andits bioavailability to marine plankton. Limnol Oceanogr 49(5):1734–1741.

27. Kirchman DL (1996) Microbial ferrous wheel. Nature 383(6598):303–304.28. Strzepek RF, et al. (2005) Spinning the “Ferrous Wheel”: The importance of the mi-

crobial community in an iron budget during the FeCycle experiment. Global Bio-geochem Cycles 19(4):GB4S26.

29. Anbar AD (2004) Iron stable isotopes: Beyond biosignatures. Earth Planet Sci Lett217(3-4):223–236.

30. Johnson CM, Beard BL (2006) Fe isotopes: An emerging technique for understandingmodern and ancient biogeochemical cycles. GSA Today 16(11):4–10.

31. Twining BS, et al. (2014) Differential remineralization of major and trace elements insinking diatoms. Limnol Oceanogr 59(3):689–704.

32. Weber L, Völker C, Schartau M, Wolf-Gladrow DA (2005) Modeling the speciation andbiogeochemistry of iron at the Bermuda Atlantic Time-series Study site. Global Bio-geochem. Cycles 19(1):GB1019.

33. Shaked Y, Kustka AB, Morel FMM, Erel Y (2004) Simultaneous determination of ironreduction and uptake by phytoplankton. Limnol Oceanogr Methods 2:137–145.

34. Barbeau K (2006) Photochemistry of organic iron(III) complexing ligands in oceanicsystems. Photochem Photobiol 82(6):1505–1516.

35. Frew RD, et al. (2006) Particulate iron dynamics during FeCycle in subantarctic waterssoutheast of New Zealand. Global Biogeochem Cycles 20(1):GB1S93.

36. Karl DM, et al. (1996) Seasonal and interannual variability in primary production andparticle flux at Station ALOHA. Deep Sea Res Part II 43(2-3):539–568.

37. Knauer GA, Martin JH, Bruland KW (1979) Fluxes of particulate carbon, nitrogen, andphosphorus in the upper water column of the northeast Pacific. Deep Sea Res Part A26(1A):97–108.

38. Martin JH, et al. (1983) Vertical Transport and Exchange of Materials in the UpperWaters of the Oceans (VERTEX): Introduction to the Program, Hydrographic Con-ditions and Major Component Fluxes During VERTEX I (Moss Landing Marine Lab,Moss Landing, CA), Tech. Publ. 83-2.

39. de Ronde CEJB, et al. (2007) Submarine hydrothermal activity along the mid-Ker-madec Arc, New Zealand: Large-scale effects on venting. Geochem Geophys Geosyst8(7):Q07007.

40. Massoth GJ, et al. (2003) Chemically rich and diverse submarine hydrothermal plumesof the southern Kermadec volcanic arc (New Zealand). Geol Soc Spec Publ 219(1):119–139.

41. Obata H, Karatani H, Nakayama E (1993) Automated determination of iron in sea-water by chelating resin concentration and chemiluminescence detection. Anal Chem65(11):1524–1528.

42. de Jong JTM, et al. (1998) Dissolved iron at subnanomolar levels in the SouthernOcean as determined by ship-board analysis. Anal Chim Acta 377(2-3):113–124.

43. Eggimann DW, Betzer PR (1976) Decomposition and analysis of refractory oceanicsuspended materials. Anal Chem 48(6):886–890.

44. Poitrasson F, Freydier R (2005) Heavy iron isotope composition of granites determinedby high resolution MC-ICP-MS. Chem Geol 222(1-2):132–147.

45. Boyle EA, et al. (2012) GEOTRACES IC1 (BATS) contamination-prone trace elementisotopes Cd, Fe, Pb, Zn, Cu, and Mo intercalibration. Limnol Oceanogr Methods 10:653–665.

46. Wessel P, Smith WHF (1998) New, improved version of generic mapping tools re-leased. Eos Trans AGU 79(47):579–579.

47. Varela DE, Pride CJ, Brzezinski MA (2004) Biological fractionation of silicon isotopes inSouthern Ocean surface waters. Global Biogeochem Cycles 18(1):GB1047.

E20 | www.pnas.org/cgi/doi/10.1073/pnas.1421576112 Ellwood et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0